LINQ is the deterministic communication network in the Qblox Cluster. It connects up to 120 Q1 sequence processors per mainframe, letting any sequencer share measurement data and conditional triggers with any other, in or across modules, with bounded latency and no host computer in the path.

Any conditional operation closes Cluster-wide in under 650 ns, a fully loaded figure covering simultaneous multi-qubit readout, all-to-all branching, and real-time parameter updates across up to 20 modules, entirely in hardware, backed by 2 Gb/s of throughput per link.

Real-time means an operation completes within a guaranteed window inside the qubit's coherence time. It's a property of determinism, not just speed. Near-time operations, like recalibration and circuit compilation, run on classical computers between circuits, on millisecond timescales. Real-time feedback has to run in hardware to hit its fixed latency budget.

QECi is the industry-standard interface for connecting control hardware to third-party decoders. LINQ supports QECi natively, so syndrome data can reach any QECi-compliant decoder without custom engineering.

Superconducting qubits, spin qubits, trapped ions, neutral atoms, and colour centres. The modular Cluster architecture lets modules combine to match a given platform, and the same LINQ feedback model applies across all of them.

 No. Nothing in the implementation is hardcoded to a single code, decoder configuration, or workflow. LINQ delivers syndrome data to any QECi-compliant decoder, and the demonstrated integration with Riverlane's Deltaflow 2 QEC system covers code distances 3 to 9.

It's the round trip time on the Surface-17 code (code distance 3), from syndrome extraction to conditional correction: all 8 ancilla qubits are read out simultaneously, the results are aggregated into a syndrome word over LINQ, that word is forwarded to the Riverlane Deltaflow 2 QEC system over QECi, and the correction is returned to the drive sequencers. Qblox's own contribution to that total is 1.114 µs and stays nearly flat as code distance increases, reaching 1.220 µs at Surface-161 (code distance 9).

Yes. Q1 sequence processors connect to NVIDIA NVQLink, extending the LINQ feedback model to GPU-accelerated compute under CUDA-Q, opening a path to GPU-based decoding and hybrid quantum-classical co-processing.

Up to 120 Q1 sequence processors per mainframe communicate over the LINQ fabric. Multiple Clusters can be daisy-chained, extending both the SYNQ timing protocol and the feedback model, with throughput scaling linearly per added Cluster.

LINQ provides 2 Gb/s of throughput per Cluster, a margin above 25 times the requirement for a Surface Code-161 at code distance 9. Throughput scales linearly as Clusters are daisy-chained, so larger codes add proportional headroom rather than eroding it.

LINQ executes at the Q1ASM assembly layer, so teams can write and iterate custom feedback logic, conditional branching, and pulse sequences directly on the sequence processors, without being limited to a fixed set of vendor-defined operations. 

Active reset and qubit frequency tracking alone measurably speed up experiment cycles today. And because LINQ already runs the full QEC-ready feedback loop in hardware, hardware bought now stays capable when a QEC roadmap becomes a near-term priority, rather than needing to be replaced.